🧠
Neuron Wiring: Solve XOR
Four points, two classes, no straight line can separate them. Hand-tune all 9 weights of this tiny 2-hidden-neuron network until the heatmap correctly colors all four corners.
🚀 Advanced
~4 min
Concept: Neural Networks · Backprop · Activations
Live decision boundary
🟠 = class 0, 🔵 = class 1. The four dots are the XOR training points — get every dot's color to match its ring.
📌 Why one layer isn't enough
A single-layer network can only draw one straight boundary. XOR needs two — this network gets there by combining an "OR-ish" neuron and an "AND-ish" neuron in its hidden layer, exactly the trick behind why deep networks stack layers.